Development of Food Inspection Device by using NIR Transmittance Imaging
INTRODUCTION
Contamination of foreign substances is one of the biggest problem for food industries. Especially in the case of organic foreign matter such as human hair and small insects are undetectable by conventional food inspection device. In this study we have developed new testing device based on NIR transmittance imaging which can detect small organic substances in food. Light diffusion equation and Monte Carlo simulation of light scattering were investigated to optimize the measurement system.
EXPERIMENTAL
The effect of incident light width on transmittance light intensity and half band width of transmittance light on flat milk chocolate was simulated by light diffusion equation. To simulate angular distribution of transmittance light, variance reduction in Monte Carlo simulation of light scattering and absorption was conducted. 50,000 photons were passed from the bottom of 10mm thickness flat milk chocolate. Fresnel reflection and refraction when photons pass across a boundary was also considered.
Based on these simulation, working prototype of in-line food inspection device composed of line scan camera with 830nm super luminescent diode. When the test samples were conveyed into the device, NIR transmittance image was automatically acquired and immediately analyzed to detect the foreign substances. Detection algorithm for small insect and human hair were developed by matlabR2014a with image processing toolbox. Detection algorithm worked in conjunction with main unit to put on a red light when foreign substances were detected. 5 mm thickness flat chocolate with human hair and insects are tested.
RESULTS AND DISCUSSION
Maximum transmittance light intensity was increased as incident light width increased. However maximum transmittance light intensity did not drastically changed over 20mm width of incident light. Half band width of transmittance light intensity increased linearly as the increment of incident light width. The ratio of transmittance intensity and half band width indicated the optimum incident light width was 7 mm. Results from Monte Carlo simulation showed that over 5mm thickness, all of transmitted photons are multiple scattered. Though chocolate is forward scattering medium, angular distribution of transmittance is almost same as calculated by isotropic scattering medium.
By using working prototype, 5 mm thickness flat chocolate with human hair and insects are investigated. In the case of chocolate with flat surface, detection algorithm worked properly. The limitation depth of human hair in chocolate was about 2 mm from the transmission surface. On the other hand, the limitation depth of small insect such as fly was about 4 mm.